Estimating hazard ratios from published Kaplan‐Meier survival curves: A methods validation study
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
OBJECTIVE: Various statistical methods have been developed to estimate hazard ratios (HRs) from published Kaplan-Meier (KM) curves for the purpose of performing meta-analyses. The objective of this study was to determine the reliability, accuracy, and precision of four commonly used methods by Guyot, Williamson, Parmar, and Hoyle and Henley. DESIGN: Pivotal randomized controlled trials (RCTs) in oncology were identified from the pan-Canadian Oncology Drug Review (pCODR) database (primary analysis) and the Food and Drug Administration's (FDA) drug approvals page (secondary analysis) between January 2012 and May 2016. Two reviewers independently reconstructed HRs using each method on KM curves extracted from each trial and compared them with reported HRs (gold standard). Bland-Altman plots and summary statistics were calculated to assess accuracy and precision of these methods. Interrater reliability was assessed using intraclass correlation coefficient (ICC). These four methods were also applied to KM curves of different structures (ie, flat versus steep curves). RESULTS: A total of 118 KM curves (55 RCTs) and 77 KM curves (46 RCTs) were extracted from pCODR and FDA, respectively. In the primary analysis, the Guyot method was the most accurate with the lowest mean error (0.0094; 95% CI, -0.0012-0.020). All four methods had excellent interrater reliability. The Guyot method showed the smallest bias and greatest precision on the Bland-Altman plots. The Guyot method was consistently superior in both the secondary and all sensitivity analyses. CONCLUSION: In the absence of reported HRs, we recommend that researchers consider the Guyot method to reconstruct HRs from KM curves when performing aggregate data meta-analyses.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.235 | 0.821 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.003 | 0.000 |
| Bibliometrics | 0.001 | 0.002 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.000 | 0.002 |
| Insufficient payload (model declined to judge) | 0.009 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it